Some of my most important work has been directed toward measuring risk aversion. This is important for many reasons; giving good advice or setting good defaults for long-term asset allocation decisions is one of them. Dan Benjamin, Mark Fontana and I explain that in the first footnote in our new NBER Working Paper “Reconsidering Risk Aversion”:

When financial advisors make their portfolio-allocation advice contingent on an individual’s risk attitudes, they typically measure the individual’s relative ranking in the population, e.g., using a qualitative scale. According to economic theory, however, what is needed is the numerical value of the individual’s risk-preference parameters (or at least the distribution of these numerical values in the population, so that an individual’s relative ranking can be interpreted numerically). These numerical values would need to be elicited using real or hypothetical choices over risky lotteries, as discussed here.

One difficulty with measuring risk aversion is that the answers people give could depend on the particular framing of a question. In “Reconsidering Risk Aversion,” we looked at whether grouping two risks together would lead to a different estimate of risk aversion than if the two risks were represented as separate risks. The surprising answer was that, while grouping or separating the risks made an idiosyncratic difference to people’s choices, it made little systematic difference in risk aversion that we could detect (in a statistical model that assumed constant relative risk aversion, expected utility maximization, and response error driven by a random risk aversion parameter). And differences in people’s choices because of framing were reduced when we gave people a chance to reconsider their choices. By contrast, when people had made corresponding choices under different framings and were given a chance to reconsider, they very seldom changed their choices.

Giving people a chance to reconsider their choices had two particularly strong effects:

• People became much more transitive in their choices

• People adjusted their choices to treat compound lotteries in a way much more similar to how they treated the corresponding simple lotteries. That is, they came much closer in accord with the “Reduction of Compound Lotteries” Axiom. (Initially, they were quite far away from being in accord with this axiom.)

There is a lot more I want to do to better understand risk aversion and its implications for life-cycle saving and asset allocation. I don’t feel done with this research agenda.